I’m pretty sure there’s a bug somewhere in my calculations for the slopes used to calculate the electric fields. I started trying to debug this and quickly realized that my current “unrolled” implementation of the slope calculations is just too unwieldy and difficult to debug. I started implementing a device function to compute the slopes, going slope by slope and tweaking it to work with each slope. I realized after working on this for awhile that it wasn’t a good way to go about it and my implementation was quickly becoming hacky. So next week I’m going to work on it from first principles and hopefully that will work better.
- Presented on my attending the International High Performance Computing Summer School